Notes

Questions

Plan of analytical attack

1. How does invert community vary with land use and veg cover?

• Visualize with NMDS
  • Point for each trap (color-coded by pitfall vs. sticky trap type)
  • Vectors for veg cover (by life form)
    • We have trap-specific vegetation data
  • Vector for land use (might need to be binary –> polygon)
• Model with multivariate GLMM (via mvabund)

taxa abundances ~ land use + veg cover (by lifeform) + RE:site + RE:trap type

• How to combine pitfall/sticky data?
  
• Option 1: As described above (for NMDS, color-code by trap type; for GLMM, include trap type as a RE)
  
• Option 2: Two separate analyses (this seems confusing)
  
• Option 3: Sum abundance by trap pair and use P/A in analyses
  
• Use relative abundance or PA?
  
• For now I’m planning to do both and compare
  
• Is it fine to use a trap as a sample (with site as a RE)?
  
• I’m worried about independence of samples, but also we only have 9 sites?
  
• Should I remove rare taxa?
  
• I’m thinking of limiting everything to the top 8(?) lizard prey taxa

2. How does lizard diet vary with habitat?

• Visualize with NMDS
  • Point for each lizard gut sample
  • Vectors for veg cover (by life form)
    • These would have to be based on site averages, since we obviously don’t have lizard-specific vegetation data
  • Vector for land use (might need to be binary –> polygon)
• Model with multivariate GLMM (via mvabund)

taxa abundances ~ land use + veg cover (by lifeform) + RE:site

Same questions as above:
- Use relative abundance or PA? (planning to do both and compare)
- Is it fine to use a lizard as a sample (with site as a RE)?
- Should I remove rare taxa? (maybe limit to top 8 prey taxa)

3. Does invert community predict lizard diet?

• Maybe visualize with some sort of correlation matrix?
  • I’m thinking prey items on the x and diet items on the y, color coded by strength of interaction?
• Model with multivariate GLMM (via mvabund)

diet abundances ~ prey abundances + RANDOM:site

• Is this even a thing you can do???
  
• I think it is I just can’t find an example with mvabund yet but haven’t looked very hard
  
• Same as above: Use relative abundance or PA?
  
• I think that

4. What about plants and ants?

• Something simple about herbivory and/or myrmecophagy because they’re especially weird/interesting
• Visualize with
  • Maybe boxplots for land use (if needs to be binary)
• Model with GLMM

herbivory ~ land use + veg cover (by lifeform) + RE:site myrmecophagy ~ land use + veg cover (by lifeform) + RE:site

Other ideas

• Diversity/abundance
• Invert biomass
• Invert functional groups (e.g. hardness, activity level)
• Prey selectivity indices

Qs 4 Colin

• Land use gradient instead of binary?

To do:

• Separate ants using comments
• Flag rare OTUs